ARTICLES. Alcohol, Low-Methionine-Low-Folate Diets, and Risk of Colon Cancer in Men

Transcription

1 ARTICLES Alcohol, Low-Methionine-Low-Folate Diets, and Risk of Colon Cancer in Men Edward Giovannucci, Eric B. Rimm, Alberto Ascherio, MeirJ. Stampfer, Graham A. Colditz, Walter C. Willett* Background: Methylation of DNA, which may have a role in the regulation of gene expression, depends on dietary folate and methionine. Because aberrant DNA methylation may contribute to the initiation or progression of colon cancer, we hypothesized that deficient intakes of folate or methionine and high consumption of alcohol, an antagonist of methyl-group metabolism, increase risk of colon neoplasia. Previously, a high-alcohol and low-methioninelow-folate (methyl-deficient) diet was shown to be related to a higher risk of colon adenomas, precursors of cancer. Purpose: Our goal was to determine if ingestion of a high-alcohol, methyl-deficient diet is related directly to risk of colon cancer. Methods: We assessed dietary intake for a 1-year period for a cohort of U.S. male health professionals, years old and free of diagnosed cancer in We assessed diet by using a validated, semiquantitative food-frequency questionnaire. During 6 years of follow-up, we documented 205 new cases of colon cancer in this cohort. Results: Current alcohol intake was directly related to risk of colon cancer (multivariate relative risk [RR] = 2.07; 95% confidence interval [CI] = , for >2 drinks versus <0.25 drink daily; P trend =.005), and past drinkers were also at higher risk (RR = 1.95; 95% CI = ). Individually, folate and methionine intakes were weakly inversely associated with risk of colon cancer. An adverse effect of a high-alcohol, low-methyl diet was not observed among regular users of aspirin, who have previously been shown to be at lower risk for colon cancer. Combinations of high alcohol and low methionine and folate intakes yielded striking associations for total colon cancer (RR = 3.30 [95% CI = ] comparing high-methyl diets with lowmethyl diets among nonusers of aspirin) and for cancers of the distal colon (RR = 7.44; 95% CI = ). Among men with high intakes of folate or methionine, alcohol levels of >2 drinks daily were not associated with risk of colon cancer. The increased risk of colon cancer associated with alcohol and methyl-deficient diets was not confounded by smoking; intakes of fat, red meat, and fiber; level of physical activity; multivitamin or aspirin use; and body mass index. Conclusions: These findings support the hypothesis that substantial consumption of alcohol, when combined with inadequate intakes of folate and methionine, may increase risk of colon cancer and confirm similar findings in adenomas. Implications: These data provide further support of recommendations to avoid excess alcohol consumption and to increase dietary folate to lower the risk of colon cancer. [J Natl Cancer Inst 87: ,1995] The methylation of DNA cytosine at specific cytosine-phospho-guanine (CpG) sites inhibits transcription and, thus, has a role in the regulation of gene expression (1). Imbalanced DNA methylation, characterized by global genomic hypomethylation (2^) and methylation of usually unmethylated CpG sites (4), is observed consistently in colonic neoplasia (5). The reduction in DNA methylation first appears in non-neoplastic colonic tissue prior to the development of the neoplasm. Hypomethylation may be initiated by inadequate cellular levels of the methyl donor S-adenosylmethionine (6). Because production of S- adenosylmethionine depends on both methionine and folate, diets deficient in these factors or high in alcohol, a methyl group antagonist (7), may cause imbalanced DNA methylation. On the basis of these relationships and supported by animal studies (8-12), we hypothesized that dietary factors that lower methyl-group availability increase risk of colon cancer in humans. This hypothesis would explain a number of epidemiologic observations, including that low intakes of fresh fruits * Affiliations of authors: E. Giovannucci, Channing Laboratory, Department of Medicine, Harvard Medical School and Brigham and Women's Hospital, Boston, Mass E. B. Rimm, A. Ascherio, Department of Epidemiology and Department of Nutrition, Harvard School of Public Health, Boston. M. J. Stampfer, W. C. Willett, Channing Laboratory, Department of Medicine, Harvard Medical School and Brigham and Women's Hospital; and Department of Epidemiology and Department of Nutrition, Harvard School of Public Health. G. A. Colditz, Channing Laboratory, Department of Medicine, Harvard Medical School and Bngham and Women's Hospital; and Department of Epidemiology, Harvard School of Public Health. Correspondence to: Edward Giovannucci, M.D., Channing Laboratory, 180 Longwood Ave., Boston, MA See "Notes" section following "References." Journal of the National Cancer Institute, Vol. 87, No. 4, February 15, 1995 ARTICLES 265

2 and vegetables (13), the major sources of folate, and high intakes of alcohol (14) appear to increase the incidence of colon cancers. In addition, although methionine has rarely been assessed directly in epidemiologic studies, many foods rich in this amino acid excluding red meat (e.g., poultry, fish, and dairy products) usually have been inversely associated with risk of colon cancer (15-18). These dietary factors also appear to be similarly related to risk of colonic adenomas (19-22), precursors of cancer. In two cohorts, we had found that risk of colorectal adenoma was elevated in individuals with high intakes of alcohol and low intakes of methionine and folate and that the excess risk was particularly high among those who had deleterious combinations, such as high alcohol low folate intake (23). In the present study, we examine these dietary patterns with colon cancer occurrence within an ongoing cohort of men. Subjects and Methods Study Population The Health Professionals Follow-up Study (24), designed to study the causes of heart disease and cancer, began in 1986 when U.S. male dentists, optometrists, osteopaths, podiatrists, pharmacists, and veterinarians years of age responded to a mailed questionnaire. Participants provided information on diet (described below), smoking history, age, weight, height, leisure-time physical activity, aspirin use, history of colonoscopy or sigmoidoscopy, parental history of colorectal cancer, and history of professionally diagnosed medicaj conditions. Current regular use of aspirin and other nonsteroidal anti-inflammatory medications was also assessed on the base-line questionnaire. The medication list included the following categories: 1) aspirin, two or more times a week (e.g., Anacin, Bufferin, or Alka-Seltzer); 2) acetaminophen, two or more times a week (e.g., Tylenol); and 3) other anti-inflammatory medications (e.g., Motnn, Indocin, Naprosyn, or Dolobid). In 1988, 1990, and 1992, we mailed follow-up questionnaires to update exposure information and to ascertain newly diagnosed medical conditions. This investigation was approved by the Harvard School of Public Health Human Subjects Committee. Dietary Data At base line (1986), we assessed dietary and alcohol intake in the cohort by using a semiquantitative food-frequency questionnaire (2526). The questionnaire included 131 food and beverage items, which accounted for more than 90% of the intake of major nutrients. For each food or beverage item, a commonly used unit or portion size was specified, and participants were asked how often on average over the past year they consumed that amount of each food. Participants chose from nine possible responses, which ranged from never to six or more times per day. We also included questions on the brand of breakfast cereal used; the brand, frequency, and duration of use of multivitamin and individual vitamin supplements; the types of fat commonly used; and an openended section for foods not listed. We computed nutrient intakes by multiplying the consumption frequency of each unit of food by the nutrient content of the specified portions, using composition values from U.S. Department of Agriculture sources (27) supplemented with other data. Before conducting any analyses, we excluded improperly completed forms, which were defined primarily by extreme values of total energy intake (outside the range of kcal/day) rather than by the number of blank food items, but we also excluded men with 70 or more blanks. Informal examination of the questionnaires suggested that forms with a high number of blanks appeared plausible and carefully completed. Furthermore, in a substudy (described below), the number of blanks was not associated with the degree of measurement error, which was defined as the absolute value of nutrient intake measured by the diet records minus nutrient intake measured by the questionnaire. Within the entire cohort, the median number of food blanks from the food list was one, and 81 % had three blanks or fewer. In 1986, we evaluated the precision of nutrient consumption measured by the questionnaire in a subsample of 127 men from the Boston area (25J26). In brief, these participants completed two 1-week dietary records and provided a fasting blood sample. The identical exclusion criteria used for the entire cohort (total caloric range and missing foods) were used in the substudy. Mean daily intakes of nutrients, including alcohol and folate, as assessed by the questionnaire and by the dietary records were very similar, and the correlation between the two methods was.86 for alcohol and.71 for folate. Also, the correlation between folate intake by the questionnaire (including supplements) and erythrocyte folate level from the blood samples was.55 (with deattenuation for within-person variability in the blood level) (23). Data on methionine intake were not available from diet records. Identification of Cases of Colorectal Cancers On each follow-up questionnaire in 1988, 1990, and 1992, we asked the participant about cancer diagnosed during the past 2 years. When a subject (or next of kin for decedents) reported a diagnosis of cancer of the colon or rectum on our follow-up questionnaire, we asked him (or next of kin) for permission to obtain hospital records and pathology reports. A study physician, blinded to exposure information, reviewed all medical records and extracted data on histologic type, anatomic location, and stage of the cancer. We confirmed 251 new cases of colorectal adenocarcinoma (excluding carcinoma in situ), 219 (87%) by medical records and 32 with corroborating information on diagnosis and treatment from the cohort member. Tumors from the cecum to the splenic flexure were considered to be in the proximal colon. Because rectal cancers may have an epidemiologic pattern different from that for colon cancers, we excluded the 46 cases of rectal cancer, leaving 205 cases of colon cancer for the primary analysis. Data Analysis Before conducting the analyses, we excluded 1530 men with implausibly high or low scores for total energy intake or with 70 or more items left blank and 2068 men who reported previous cancer (other than nonmelanoma skin cancer), ulcerative colitis, or a familial polyposis syndrome. The 2025 men who reported previous cancers were excluded because those with prevalent malignancies may have altered their diets for psychological and patho-physiologic reasons (e.g., related to tumor burden); thus, their base-line assessment may not reflect longterm dietary intake. The remaining men contributed to follow-up time beginning with the return of the initial questionnaire in 1986 and ending at the month of diagnosis of colorectal cancer, month of death from other causes, or at the end of the study period, January 31,1992. The response rate to the follow-up questionnaires was 94% through January 31, Most of the deaths in the cohort were reported by family members or the postal system in response to the follow-up questionnaires. In addition, we used the National Death Index, a highly sensitive method of identifying deaths among nonrespondents (28). We computed incidence rates for each category by dividing the number of incident cases by the number of person-years in that category and computed the relative risk (RR) by dividing the incidence rate in each category by the rate in the specified reference group. We considered potentially confounding effects of other variables believed to be important regarding risk of colon cancer. Included among the factors we examined were lifetime cigarette smoking; current physical activity level; body mass index; aspirin use; multivitamin use; total energy, fat, red meat, vitamin D, and calcium intake; history of polyps; and history of endoscopic screening. We used the Mantel Haenszel summary estimator and proportional hazards modeling to adjust for age (across 5-year categories) and other potentially confounding variables. Tests for statistical interaction were conducted by including a crossproduct term of the two variables of interest in a proportional hazards model. All P values presented are for two-sided tests. Categories of nutrients were determined by quintiles, except for alcoholic beverages which were categorized into levels of drinks per day (beer one glass, bottle, or can; red wine or white wine one 4-oz. glass; liquor one drink or shot). The alcohol content was estimated to be 13.2 g for a 12-ounce bottle or a 12-ounce can of beer, 10.8 g for a glass of wine, and 15.1 g for a drink of liquor. Nondrinkers were defined as men who did not drink currently and who did not report (on the base-line 1986 questionnaire) that their drinking had significantly decreased during the past 10 years. Current nondrinkers who reported that their drinking had decreased substantially were classified separately as past drinkers. 266 ARTICLES Journal of the National Cancer Institute, Vol. 87, No. 4, February 15, 1995

3 Folate and methionine intakes were adjusted for total energy intake using residual analysis (29). We considered the independent and the joint effects of folate, methionine, and alcohol intakes, and we used combinations of these factors to characterize high-methyl and low-methyl dietary patterns. To be consistent with our previous article on adenomas from this cohort (23), the categorizations were based on extreme quintiles for folate and methionine and on less than 5 g/day and greater than 20 g/day for alcohol. Results We first considered the effect of alcohol, methionine, and folate independently of each other. Current and past intakes of alcohol were related to increased risk of colon cancer, and level of current consumption displayed a dose-response relationship. Because few cases of colon cancer occurred among nondrinkers and because the incidence rate was similar among light drinkers of drink per day, these two groups were combined to form a more statistically stable reference group (Table 1). In the multivariate model, the RRs were similar except for a slight attenuation in the top category (>2 drinks daily). The increased risk associated with alcohol intake was stronger in the distal colon but was also elevated, though not significantly, in the proximal colon. The incidence of colon cancer rose progressively with increasing consumption of all three alcoholic beverages, which was suggestive of an effect of ethanol rather than of any specific beverage type. The RR and 95% confidence intervals (CIs) comparing 15 g or more of alcohol content per day versus 0 g/day was 1.35 (95% CI = ) for beer, 2.06 ( ) for wine, and 1.56 ( ) for liquor (from the multivariate model described in Table 1 except that it included alcohol from each of the beverage types simultaneously rather than total alcohol). Only the higher risk associated with wine intake was statistically significant (P trend =.03), but the 95% CIs for all beverages were wide and overlapped substantially. Men at the lowest level of methionine intake were at elevated risk of developing colon cancer (multivariate RR [95% CI] from the low to high quintiles: 1.0 [ ]; 0.71 [ ]; 0.84 [ ]; 0.82 [ ]; and 0.65 [ ]). However, the test for overall trend across quintiles was not statistically significant (P =.31). An inverse association was suggested for the distal colon (multivariate RR = 0.53; 95% CI = , for extreme categories; P trend =.12) but not for the proximal colon (multivariate RR = 0.91; 95% CI = ; P trend =.64). Total folate intake and folate from foods only were not significantly related to risk of colon cancer (total folate: multivariate RR = 0.86; 95% CI = ; P trend =.30). Because folate intake may act during the transition of normal mucosa to adenoma (25) and the transition from non-neoplastic mucosa to invasive carcinoma may take 10 years or longer, we examined duration of use of folate supplementation (primarily in multivitamins) to estimate the more etiologically relevant exposure. Men who had used multivitamins containing folate were at reduced risk compared with the risk for nonusers (for 1-9 years of use, multivariate RR [95% CI = ]; for 10 years of use, multivariate RR = 0.74 [95% CI = ]), although the difference was not statistically significant. Longterm use was a stronger risk factor for cancers of the distal colon (1-9 years: RR = 0.97 [95% CI = ]; >10 years: RR = 0.62 [95% CI = ]) than for cancers of the proximal colon. Because the correlation between folate and vitamin D from supplements was high (r =.83), we could not discriminate between these two nutrients. We next examined whether the effects of alcohol, methionine, and folate intake varied across possibly modifiable risk factors, including other nutrients, physical activity, smoking, body mass index, and aspirin use. We found that the influence of alcohol and methionine on colon cancer risk was stronger among men who did not use aspirin regularly (Table 2). Aspirin use was shown previously to be inversely associated with risk of colorectal cancer in this cohort (30). The P value for the test for statistical interaction between aspirin use and alcohol intake was.07; for the test for statistical interaction between aspirin use and methionine intake, the P value was.11. Aspirin use did not appear to be as strong a modifier for folate intake. None of the other factors we considered appeared to modify the risk due to Table 1. RR of colon cancer by level of alcoholic beverage intake in men, (reference category) Average drinks per day >2.0 Past drinkers P trend* Person-years Total colon cancer, No. of cases Age-adjusted RR (95% CI) Multivariate RRf (95% CI) Distal colon cancer.j No. of cases Age-adjusted RR (95% CI) Multivariate RRf (95% CI) Proximal colon cancer,:): No. of cases Age-adjusted RR (95% CI) Multivariate RRf (95% CI) ( ) 1.42( ) ( ) 1.69( ) ( ) 0.68 ( ) ( ) 1.67( ) ( ) 2.80( ) ( ) 1.08( ) ( ) 1.68( ) ( ) 2.51 ( ) ( ) 0.89 ( ) ( ) 2.07( ) ( ) 2.01 ( ) ( ) 2.23( ) ( ) 1.95( ) (O.85^».O3) 1.81 ( ) ( ) 2.53( ) Test for trend determined by modeling drinks per day as a continuous variable in the multivariate model (excluding past drinkers). trr adjusted by proportional hazards modeling for age, history of polyps, history of endoscopy, parental history of colorectal cancer, total pack-years of cigarettes, body mass index, leisure-time physical activity, aspirin use, and intakes of energy, red meat, methionine, and folate. JThe sum of distal and proximal tumors does not add up to total because of missing data on subsites. Journal of the National Cancer Institute, Vol. 87, No. 4, February 15, 1995 ARTICLES 267

4 Table 2, RR of colon cancer by intake of alcohol, methionine, and folate among aspirin users and nonusers Alcohol, dnnks per day (reference category) >2.0 Past dnnker P trend Aspirin, nonusers: RR* (95% CI) Aspirin, users: RR* (95% CI) 1.49( ) 1.30( ) 1.63( ) 1.68( ) 1.77( ) 1.47( ) 2.52( ) 1.38( ) 2.47( ) 0.92 ( ).003t.62f Methionine, g/day < >2.44 /"trend Aspirin, nonusers: RR* (95% CI) Aspirin, users: RR* (95% CI) 10( ) 0.75( ) 0.62( ) 0.78( ) 1.07( ) 0.75( ) 1.11( ) 0.56 ( ) 0.% ( ) Folate, )ig/day Aspirin, nonusers: RR* (95% CI) Aspirin, users: RR* (95% CI) < ( ) 1.16( ) ( ) 1.83( ) ( ) 0.79( ) > ( ) 0.82 ( ) *From multivariate model including age, history of polyps, history of endoscopy, parental history of colorectal cancer, total pack-years of cigarettes, body mass index, leisure-time physical activity, and intake of total energy, red meat, methionine, and folate. hi total, there were 149 cases among nonusers of aspirin and 56 cases among aspirin users. t Excluding past drinkers. Low alcohol intake, 220 g/day RR* (95% CI) High alcohol intake, >20 g/day RR* (95% CI) P trend methyl-group availability, and use of other nonsteroidal anti-inflammatory drugs was too low to evaluate. We conducted further analyses of methyl-related factors both for the total cohort and for nonusers of aspirin. Because of the suppressive effects of alcohol on methylgroup metabolism, we suspected that low intakes of methionine or folate might be stronger risk factors for colon cancer among substantial alcohol users. Associations between intakes of methionine and folate and risk of colon cancer were essentially limited to substantial consumers of alcohol or past drinkers, presumably mostly heavy drinkers (Table 3). Among men in the high quintile of methionine intake, consumption of alcohol did not appear to increase risk, but as intake of methionine decreased, higher alcohol consumption was related to higher risk of colon cancer. Among substantial alcohol consumers excluding aspirin users, the risk of colon cancer increased progressively across decreasing intakes of methionine (P trend =.02), and a particularly high risk was seen in the bottom versus the top quintiles (RR = 5.0; 95% CI = ). Alcohol consumption did not raise the risk among men in the highest quintile of Table 3. RR of colon cancer by level of methionine or folate intakes and alcohol level < / ( ) 33/ ( ) / ( ) 18/ ( ) Methonine, g/day / ( ) 15/ ( ) Folate, ig/day / ( ) 17/ ( ) > / (ref)t 10/ ( ) < >646 Low alcohol intake, ^20 g/day RR* (95% CI) High alcohol intake, >20 g/day RR* (95% CI) 19/ ( ) 21/ ( ) 23/ ( ) 22/ ( ) 21/ ( ) 19/ ( ) 27/ ( ) 18/ ( ) 22/ (ref)t 13/ ( ) *From multivariate model including age, history of polyps, history of endoscopy, parental history of colorectal cancer, total pack-years of cigarettes, body mass index, leisure-time physical activity, total energy and red meat intakes, as well as the specified combinations of folate, meltuonine, and alcohol. treference category. 268 ARTICLES Journal of the National Cancer Institute, Vol. 87, No. 4, February 15, 1995

5 folate intake; most of these men were taking supplements. In the four lower quintiles of folate intake, higher alcohol intake was associated with higher risk. Statistical tests for interaction with alcohol intake yielded P values of.04 for methionine and.34 for folate intake. To examine extreme dietary contrasts of methyl-group availability, we combined intakes of alcohol, folate, and methionine as defined in a previous analysis (23). Age distribution did not differ across levels of methyl-group availability (low-methyl diet: mean age in 1986 = 54.9 years, and standard deviation [SD] = 9.9; high-methyl diet: mean age = 54.8 years, and SD = 10.0). By definition, a high-methyl diet was characterized by low alcohol consumption, approximately twice the frequency of consumption of methionine-rich foods (poultry, fish, and low-fat dairy products), and higher intake of folate sources, including multivitamin supplements and vegetables (Table 4). Factors related to a "Western" lifestyle (i.e., high red meat intake, physical inactivity, increased body mass index, and high percent of energy from fat [excluding energy from alcohol]) that are hypothesized to increase risk of colon cancer did not vary appreciably among the groups defined by methyl availability of the diet. Low-methyl diets tended to be lower in poultry, fish, low-fat dairy products, fruits, and vegetables and higher in alcohol, sweets and other highly refined and processed foods, and high-fat dairy products. Men on a low-methyl diet (high alcohol-low methionine-low folate) were at markedly higher risk of developing colon cancer than men on a high-methyl diet, with particularly strong associations in the distal colon and among nonusers of aspirin (RR = 7.44; 95% CI = ) (Table 5). Since the methyl-poor diet and the Western lifestyle patterns were almost entirely independent (see Table 4), the increased risk of colon cancer associated with methyl-deficient diets was not confounded by intakes of fat and red meat, physical activity, and body mass index. Also, other dietary factors, including calcium, vitamin D, Folate, mg/dayf Methionine, g/day Alcohol, g/day Past heavy drinkers, % Body mass index, kg/m 2 Physical activity, metabolic equivalents/wk Aspirin use, % Beef, pork, or lamb as a main dish, servings/day Total energy, kcal/dayt Calories from fat, % Poultry, servings/day Fish, servings/day High-fat dajry products, servings/day Low-fat dairy products, servings/day Vegetables, servings/day Fruits, servings/day Grains and baked products, servings/day Multivitamin use, % Highly refined sweets, including beverages, servings/day and fiber, were not confounders. Multivariate analyses among nonusers of aspirin based on the extreme contrasts of alcohol and folate or methionine in pairs yielded analogous results, though the 95% CIs were wide (alcohol-methionine: RR = 7.8, 95% CI = , and P =.02; alcohol-folate: RR = 10.4, 95% CI= , and P =.01). In our previous analyses of adenomas (23), we found particularly strong relations between high intake of alcohol and low intake of folate and methionine and risk of colon adenoma among younger men; thus, we examined these associations among younger men. Among men younger than 65 years of age at base line in the high-methyl category, no distal colon cancers were documented in person-years, with 11.3 cases expected on the basis of age-adjusted rates of the methyl-deficient group; among these men on a high-methyl diet, six total colon cancer cases were documented versus 20.3 expected. These results suggest that, even within a relatively high risk Western population, cancers of the distal colon are rare among those on a high-methyl diet. Discussion Table 4. Characteristics of diets defined by combinations of methyl-group factors* High methyl (n = 4687) Methyl-deficient diets, probably by causing DNA methylation abnormalities, cause various cancers in animals (8-12). Because similar DNA methylation aberrations occur in human colon neoplasia (5), we hypothesized that low intakes of folate and methionine as well as substantial consumption of alcohol increase risk of colon cancer. Results from this cohort study of men supported this hypothesis. We found that substantial alcohol consumption in concert with low intakes of folate and methionine was associated with an increased risk of colon cancer, especially carcinomas of the distal colon. The associations observed for alcohol and methionine were not due to confounding by other dietary factors, smoking, physical activity, body mass, aspirin use, differential surveillance for disease, or family Intermediate (n = ) Low methyl (n = 6149) *High methyl defined as alcohol drinkers of <5 g/day and either energy-adjusted folate or methionine intake in highest quintile, and low methyl defined as alcohol drinkers >20 g or past heavy drinkers plus either folate or methionine intake in the lowest quintiles. texcluding calories from alcohol. Journal of the National Cancer Institute, Vol. 87, No. 4, February 15, 1995 ARTICLES 269

6 Table 5. RR of colon cancer by site according to combination of alcohol, methionine, and folate intake* Low alcohol-high methionine-high folate Intermediate High alcohol-low methionine-low folate Total colon cancer RR (95% CRt Distal colon cancerll RR (95% CI) Proximal colon cancer RR (95% CI) Total colon cancer RR (95% CI)# Distal colon cancer RR (95% CI) Proximal colon cancer RR (95% CI) 14/ t 4/ * 7/ (» 9/ (» 2/ (-)t 4/ (» Total 145/ ( ) 62/ ( ) 44/ ( ) Excluding aspirin users 101/ ( ) 43/ ( ) 32/ ( ) 46/ (1.28^.30) 23/ ( )1 18/ ( ) 39/ ( )5 20/ ( ) 14/ ( )** High alcohol is defined as >20 g/day and low alcohol as <5 g/day; high and low energy-adjusted folate and methionine are defined by high and low quintiles of intake {see Table 4). The intermediale group is comprised of individuals who do not fall into the extreme combinations of alcohol, methionine, and folate. frr adjusted for age, history of polyps, history of endoscopy, total pack-years of smoking, body mass index, leisure-time physical activity, red meat consumption, total energy intake, aspnn use, and multivitamin use. ^Reference category. IIThe sum of distal and proximal tumors does not add up to total because of missing data on subsites. V01S/><.05. #Adjusted for same covariates except for aspirin use. **.05</><.10. history of colorectal cancer. For folate, we can be less certain of controlling for confounding because of folate's high correlation with other nutrients in multivitamins, particularly vitamin D. However, in a previous analysis of colonic adenomas involving 895 cases (23), an inverse association with folate intake persisted even after controlling for intakes of fiber and vitamins A, C, D, and E. The degree of internal consistency within our data strengthened our confidence in the results. Three separate dietary factors (folate, methionine, and alcohol) contributed to this risk; for all three factors, the association was strongest in the distal colon. The influence of folate and methionine was limited to substantial consumers of alcohol, which is consistent with a suppressive effect of alcohol on methyl-group metabolism, and alcohol consumption did not appear to confer excess risk if the intake of methionine or folate was high. All types of alcoholic beverage were related to occurrence of colon cancer, and past, presumably heavy, drinkers were also at higher risk of developing colon cancer. Moreover, the same factors were similarly related to risk of colonic adenomas in this and in another cohort of women (23). A factor that we have failed to consider is unlikely to account for all these observations, especially given the strength of the associations (e.g., RR exceeding 10 for high alcohol-low folate versus low alcohol-high folate). By focusing our study on overall methyl-group content of the diet, we were able to avoid some statistical problems related to collinearity that may arise when examining single nutrients individually. For example, folate intake is correlated with multivitamin use, and methionine is correlated with protein intake; thus, it may be difficult statistically to differentiate folate intake from some other component of multivitamin use or methionine from protein. However, our definition of methyl-group content of diet includes the combined influence of folate, alcohol, and methionine, and this overall methyl-group variable is much less inherently correlated with potentially confounding factors as would be some of its constituents (e.g., as multivitamin use would be with folate). Thus, while it may be problematic to include folate and multivitamin use in the same multivariate model, when we included multivitamin use in the same model with the overall methyl-group content of the diet (Table 5), multivitamin use was not a statistically significant factor. Moreover, the coefficients and 95% CIs for methyl-deficient diets were nearly identical with and without multivitamin use in the multivariate model. Stated another way, our analyses suggest that multivitamin use per se did not contribute to risk of colon cancer independently to its contribution to the overall methyl-group availability of the diet. The proposed mechanism whereby methyl-deficient diets influence cancer is best understood from studies of rat hepatocarcinogenesis. That rats maintained on a methyl-deficient diet develop hepatocellular carcinomas was first reported in 1946 (31). It is remarkable that this was the first report in which omission of a substance from the diet resulted in neoplasia. More recently, it has been shown that the introduction of a methyl- 270 ARTICLES Journal of the National Cancer Institute, Vol. 87, No. 4, February 15, 1995

7 deficient diet to rats is followed by DNA hypomethylation, the overexpression of various genes including the proto-oncogenes c-fos, c-ha-ras, and c-myc, and an elevated DNA methyltransferase activity in the liver (5,9). In the short term, these changes are reversible, but animals on a long-term methyl-deficient diet develop liver tumors (72). Alcohol appears to accentuate this effect (32). In this model, the promoting effect of a methyl-deficient diet appears to be mediated directly through abnormalities in DNA methylation (10,33). More recently, a moderate folate deficiency has also been linked to enhanced dimethylhydrazine-induced colon dysplasia and neoplasia in rats (8), and recent evidence suggests that a high-folate supplement reverses colonic DNA hypomethylation in humans (34). Abnormal DNA methylation patterns may contribute to carcinogenesis, possibly by influencing both the activation of oncogenes and the inactivation of tumor suppressor genes (2-5). Initially, the reduction in genomic methylation caused by methyl-deficient diets may contribute to the loss of the normal controls on proto-oncogene expression. Yet, in the midst of overall hypomethylation (2,3), specific CpG sites may become abnormally methylated (4), probably as a result of elevated DNA methyltransferase activity that appears to follow a methylgroup deficiency. Of potential relevance, a CpG island-rich area on chromosome 17p, normally unmethylated in adults, is extensively hypermethylated in colonic tumors. Over progressive tumor stages, the degree of hypermethylation on this region increases and precedes the characteristic allelic losses of this chromosome in colon cancers (4). Given the stronger dietary associations for distal colon cancers, it is noteworthy that allelic losses occur more frequently in distal cancers (35). Although an inverse association between folate and risk of colon cancer has been reported in only a few studies (36 J7), many studies (38-46) have found an inverse association between intake of fresh vegetables, the major sources of folate, and risk of colonic neoplasia. The study by Freudenheim et al. (36) also reported that men with simultaneously high alcohol and low folate intakes were at particularly high risk of developing colorectal cancer. Methionine has also received little direct epidemiologic investigation, but many animal foods high in this amino acid other than red meat (including fish, poultry, and milk products), as well as vegetable protein, are frequently inversely associated with risk of colon cancer in cohort studies and with risk of adenoma (15-22). A protective influence of methionine may at first seem somewhat paradoxical given that many case-control studies (43,44,46-50) and some recent cohort studies (17J8J1) have indicated that some component of red meat intake may increase risk of colon cancer. However, red meat is only one of several contributors to methionine, and a previous analysis from this cohort indicated that whatever factor in red meat accounts for its frequent direct association with colon cancer, it is unlikely to be its protein or methionine content (18). It is also important to note that, in this population at least, red meat intake was not appreciably correlated with the overall methyl-group status of the diet (see Table 4). We speculate that other constituents of red meat may be deleterious. For example, high consumption of red meat may increase concentrations of fecal iron, which could influence risk of colon cancer via the generation of hydroxyl radicals (52-55), or, alternatively, carcinogens (56) and promoters formed when meat is cooked (57) may be critical. Other possible explanations exist, as summarized previously (18). Our results showing an association between alcohol intake and colon cancer risk are consistent with many ecological (14), cohort (58-66), and population-based case-control (38,67-73) studies. Furthermore, alcohol consumption is rather consistently related to higher risk of colorectal adenoma (2023,74-80). An extensive review by Kune and Vitetta (14) concluded that a positive association between alcohol intake and colorectal cancer was found in five of seven correlational studies, in nine of 10 studies using community controls, but in only five of 17 studies that used hospital-based controls. These authors suggested that some of the hospital-based studies may be biased because alcohol intake is related to many conditions requiring hospitalization (81), which could cause an overestimate of intake among controls. They also found that of 14 cohort studies, an association with alcohol was noted in 10, while in three of the four cohort studies that did not demonstrate an t association, the alcohol data collected were limited. Studies have found most consistently that the elevated risk associated with alcohol occurs predominantly in the rectum or distal colon (14), whereas fewer studies report an elevated risk associated with proximal colon cancer. Because high intakes of folate or methionine appeared to mitigate the influence of alcohol, we hypothesize that varying dietary patterns among populations may modify the risk associated with alcohol and thus may contribute to the apparent inconsistencies among studies. Beyond alcohol, folate, and methionine, other dietary components such as vitamin B 12 and choline may relate to methylgroup availability. In one study (52), serum vitamin B 2 deficiency was more strongly predictive than folate of subnormal fasting levels of serum methionine. We do not consider dietary vitamin B, 2 here because in a reasonably well nourished population, low serum vitamin B ]2 is caused almost entirely by poor intestinal absorption rather than by inadequate dietary intake (83). Because one small study (84) suggested an elevated risk of colorectal cancer among patients with pernicious anemia (vitamin B 12 deficiency), an investigation relating vitamin Bj 2 levels to colon cancer risk is warranted. A potential role of choline intake in the ultimate production of 5-adenosylmethionine is unclear because choline can be produced in the human body from pathways both requiring and not requiring methionine, and choline is widespread in the food supply. To our knowledge, no choline deficiency has ever been produced in humans by an absence of choline in the diet (excluding total parenteral nutrition) (85). Rats are much more susceptible to choline deficiency than are humans, perhaps accounting for why the earliest indications of the effects of methyl-deficient diets were based on choline-deficient diets among rats (31). Aspirin use appeared to modify the risk associated with alcohol and methionine intake. Though unanticipated, this modification of risk was somewhat striking because it involved two of the methyl-group factors and because of its magnitude (see Table 2). This modification of risk by aspirin requires confirmation in other populations. One possible explanation is that methylation, abnormalities are more likely to progress in rapidly replicating tumor cells among individuals with methyl-deficient Journal of the National Cancer Institute, Vol. 87, No. 4, February 15, 1995 ARTICLES 271

8 diets, and aspirin may reduce the rate of cell proliferation (86) and thus attenuate the effect of a methyl-deficient diet. However, the mechanism of the apparent aspirin effect on colon cancer is not established (86), and other explanations are plausible. Our findings suggest that a diet based on a generous intake of fresh fruits, vegetables, and whole-grain foods, the replacement of poultry, fish, and low-fat dairy products for red meat, and the avoidance of excessive alcohol may help lower risk of colon cancer. This dietary plan is consistent with that recommended by various health agencies (87). Our study does not support the use of methionine supplements, particularly because excessive methionine intake can increase blood homocysteine levels, a risk factor for coronary heart disease (88). While multivitamin supplements should not be used as substitutes for a diet abundant in fruits and vegetables, their possible use as adjuncts, especially in some groups prone to low folate intake (e.g., the elderly or regular consumers of alcohol), is worth consideration. In conclusion, we found that a methyl-deficient diet, particularly in combination with substantial alcohol consumption, is associated with an increased risk of colon cancer. This elevated risk is independent of other risk factors of colon cancer usually associated with economic development or Westernization, such as physical inactivity and high fat or meat and low fiber consumption. Our findings also suggest that alcohol consumption in moderation and in conjunction with high folate intakes may not substantially increase risk of colon cancer. Future research should be aimed at determining whether the accumulation of DNA methylation abnormalities is a prime mover of the progression of colon cancers and whether these abnormalities can be limited or reversed by diet. References (/) Holliday R: The inheritance of epigenetic defects. Science 238: , 1987 (2) Feinberg AP, Vogelstein B: Hypomethylation distinguishes genes of some human cancers from their normal counterparts. Nature 301:89-92, 1983 (3) Goelz SE, Vogelstein B, Hamilton SR, et al: Hypomethylation of DNA from benign and malignant human colon neoplasms. Science 228: , 1985 (4) Makos M, Nelkin BD, Lerman MI, et al: Distinct hypermethylation patterns occur at altered chromosome loci in human lung and colon cancer. Proc Natl Acad Sci U S A 89: , 1992 (5) Hoffman RM: Altered methionine metabolism, DNA methylation and oncogene expression in carcinogenesis. A review and synthesis. Biochim Biophys Acta 738:49-87, 1984 (6) Cooper AJ: Biochemistry of sulfur-containing amino acids. Annu Rev Biochem 52: , 1983 (7) Finkelstein JD, Cello JP, Kyle WE: Ethanol-induced changes in methionine metabolism in rat liver. Biochem Biophys Res Commun 61: ,1974 (8) Shivapurkar N, Poirer LA: Tissue levels of S-adenosylmethionine and S- adenosylhomocysteine in rats fed methyl-deficient diets for one to five weeks. Carcinogenesis4: , 1983 (9) Wainfan E, Dizik M, Stender M, et al: Rapid appearance of hypomethylated DNA in livers of rats fed cancer-promoting, methyl-deficient diets. Cancer Res 49: , 1989 (10) Pascale R, Simile MM, Ruggiu ME, et al: Reversal by 5-azacytidine of the S-adenosyl-L-methionine-induced inhibition of the development of putative preneoplastic foci in rat liver carcinogenesis. Cancer Lett 56: , 1991 (//) Cravo ML, Mason JB, Dayal Y, et aj: Folate deficiency enhances the development of colonic neoplasia in dimethylhydrazine-treated rats. Cancer Res 52: , 1992 (12) Newbeme PM, Rogers AE: Lipotropic factors and carcinogenesis. In Human Nutrition: Cancer and Nutrition (Alfin-Slater RB, Kritchevsky D, eds). New York: Plenum Press, 1991, pp (13) Widen WC: The search for the causes of breast and colon cancer. Nature 338: ,1989 (14) Kune GA, Vitetta L: Alcohol consumption and the etiology of colorectal cancer a review of the scientific evidence from 1957 to Nutr Cancer 18:97-111, 1992 (15) Stemmermann GN, Nomura AM, Heilbrun LK: Dietary fat and the risk of colorectal cancer. Cancer Res 44: , 1984 (16) Hirayama T: A large-scale study on cancer risks by diet with special reference to the risk reducing effects of green-yellow vegetable consumption. In Diet, Nutrition, and Cancer (Hayashi Y, Magao M, Sugimura T, et al, eds). Tokyo: Japan Scientific Societies Press, 1986, pp (17) Willett WC, Stampfer MJ, Colditz GA, et al: Relation of meat, fat, and fiber intake to the risk of colon cancer in a prospective study among women [see comment citations in Medline]. N Engl J Med 323: , 1990 (IS) Giovannucci E, Rimm EB, Stampfer MJ, et al: Intake of fat, meat, and fiber in relation to risk of colon cancer in men. Cancer Res 54: , 1994 (19) Neugut Al, Garbowski GC, Lee WC, et al: Dietary risk factors for the incidence and recurrence of colorectal adenomatous polyps. A case-control study [see comment citation in Medline]. Ann Intern Med 11:91-95, 1993 (20) Kato I, Tominaga S, Matsuura A, et al: A comparative case-control study of colorectal cancer and adenoma. Jpn J Cancer Res 81: , 1990 (2/) Little J, Logan RF, Hawtin PG, et al: Colorectal adenomas and diet: a casecontrol study of subjects participating in the Nottingham faecal occult blood screening programme. BrJ Cancer67: , 1993 (22) Sandier RS, Lyles CM, Peipins LA, et al: Diet and risk of colorectal adenomas: macronutrients, cholesterol, and fiber [see comment citation in Medline]. J Natl Cancer Inst 85: , 1993 (23) Giovannucci E, Stampfer MJ, Colditz GA, et al: Folate, methionine, and alcohol intake and risk of colorectal adenoma [see comment citation in Medline]. J Natl Cancer Inst 85: , 1993 (24) Rimm EB, Giovannucci EL, Willett WC, et al: Prospective study of alcohol consumption and risk of coronary disease in men [see comment citation in Medline]. Lancet 338: , 1991 (25) Rimm EB, Giovannucci EL, Stampfer MJ, et al: Reproducibility and validity of an expanded self-administered semiquantitative food frequency questionnaire among male health professionals. Am J Epidemiol 135: , 1992 (26) Giovannucci E, Colditz G, Stampfer MJ, et al: The assessment of alcohol consumption by a simple self-administered questionnaire. Am J Epidemiol 133: ,1991 (27) US Department of Agriculture: Agricultural Handbook No. 8 series. Composition of Foods: Raw, Processed and Prepared. Washington, DC: US Govt Print Off, (28) Stampfer MJ, Willett WC, Speizer FE, et al: Test of the National Death Index. Am J Epidemiol 119: , 1984 (29) Willett WC, Stampfer MJ: Total energy intake: implications for epidemiologic analyses. Am J Epidemiol 124:17-27, 1986 (30) Giovannucci E, Rimm EB, Stampfer MJ, et al: Aspirin use and risk for colorectal cancer and adenoma in male health professionals [see comment citation in Medline]. Ann Intern Med 121: , 1994 (31) Copeland DH, Salmon WD: The occurrence of neoplasms in the liver, lungs and other tissues of rats as a result of prolonged choline deficiency. AmJPathol 22: , 1946 (52) Porta EA, Markell N, Dorado RD: Chronic alcoholism enhances hepatocarcinogenicity of diethylnitrosamine in rats fed a marginally methyl-deficient diet. Hepatology 5: , 1985 (33) Pascale RM, Simile MM, Sana G, et al: Comparative effects of L- methionine, S-adenosyl-L-methionine and 5'-methylthioadenosine on the growth of preneoplastic lesions and DNA methylation in rat liver during the early stages of hepatocarcinogenesis. Anticancer Res 11: , 1991 (34) Cravo M, Fidalgo P, Pereira AD, et al: Folate supplementation increases the degree of DNA methylation in patients with colonic neoplasms. Gastroenterology 104:A394, 1993 (35) Delattre O, Olschwang S, Law DJ, et al: Multiple genetic alterations in distal and proximal colorectal cancer. Lancet 2: , 1989 (36) Freudenheim JL, Graham S, Marshall JR, et al: Folate intake and carcinogenesis of the colon and rectum. Int J Epidemiol 20: , 1991 (37) Benito E, Stiggelbout A, Bosch FX, et al: Nutritional factors in colorectal cancer risk: a case-control study in Majorca. Int J Cancer 49: , 1991 (38) Potter JD, McMichael AJ: Diet and cancer of the colon and rectum: a casecontrol study. J Natl Cancer Inst 76: , 1986 (39) Macquart-Moulin G, Riboli E, Comee J, et al: Case-control study on colorectal cancer and diet in Marseilles. Int J Cancer 38: ,1986 (40) Modan B, Barell V, Lubin F, et al: Low-fiber intake as an etiologic factor in cancer of the colon. J Natl Cancer Inst 55:15-18, ARTICLES Journal of the National Cancer Institute, Vol. 87, No. 4, February 15, 1995

9 (41) Martinez I, Torres R, Frias Z, et al: Factors associated with adenocarcinoma of the large bowel in Puerto Rico. In Advances in Medical Oncology, Research and Education (Birch JM, ed), vol 3. New York: Pergamon Press, 1979, pp (42) Graham S, Dayal H, Swanson M, et al: Diet in the epidemiology of cancer of the colon and rectum. J Natl Cancer Inst 61: , 1978 (43) Manousos O, Day NE, Trichopoulos D, et al: Diet and colorectal cancer: a case-control study in Greece. Int J Cancer 32:1-5, 1983 (44) La Vecchia C, Negri E, Decarli A, et al: A case-control study of diet and colo-rectal cancer in northern Italy, hit J Cancer 41: , 1988 (45) Slattery ML, Schumacher MC, Smith KR, et al: Physical activity, diet, and risk of colon cancer in Utah [see comment citation in Medline]. Am J Epidemiol 128: ,1988 (46) Miller AB, Howe GR, Jain M, et al: Food items and food groups as risk factors in a case-control study of diet and colo-rectal cancer, hit J Cancer 32: ,1983 (47) Gerhardsson de Verdier M, Hagman U, Peters RK, et al: Meat, cooking methods and colorectal cancer: a case-referent study in Stockholm. Int J Cancer 49: , 1991 (48) Young TB, Wolf DA: Case-control study of proximal and distal colon cancer and diet in Wisconsin. Int J Cancer 42: , 1988 (49) Benito E, Obrador A, Stiggelbout A, et al: A population-based case-control study of colorectal cancer in Majorca. I. Dietary factors, hit J Cancer 45:69-75, 1990 (50) Lee HP, Gourley L, Duffy SW, et al: Colorectal cancer and diet in an Asian population a case-control study among Singapore Chinese. Int J Cancer43: , 1989 (5/) Goldbohm RA, van den Brandt PA, van't Veer P, et al: A prospective cohort study on the relation between meat consumption and the risk of colon cancer. Cancer Res 54: , 1994 (52) Babbs CF: Free radicals and the etiology of colon cancer. Free Radic Biol Med8: , 1990 (55) Younes M, Trepkau HD, Siegers CP: Enhancement by dietary iron of lipid peroxidation in mouse colon. Res Commun Chem Pathol Pharmacol 70: ,1990 (54) Siegers CP, Bumann D, Barctton G, et al: Dietary iron enhances the tumor rate in dimethylhydrazine-induced colon carcinogenesis in mice. Cancer Lett 41: , 1988 (55) Nelson RL, Yoo SJ, Tanure JC, et al: The effect of iron on experimental colorectal carcinogenesis. Anticancer Res 9: , 1989 (56) Ames BN: Dietary carcinogens and anticarcinogens. Oxygen radicals and degenerative diseases. Science 221: , 1983 (57) Zhang XM, Stamp D, Minkin S, et al: Promotion of aberrant crypt foci and cancer in rat colon by thermolyzed protein. J Natl Cancer Inst 84: ,1992 (58) Pollack ES, Nomura AM, Heilbrun LK, et al: Prospective study of alcohol consumption and cancer. N Engl J Med 310: , 1984 (59) Klatsky AL, Armstrong MA, Friedman GD, et al: The relations of alcoholic beverage use to colon and rectal cancer [see comment citation in Medline]. Am J Epidemiol 128: , 1988 (60) Wu AH, Paganini-Hill A, Ross RK, et al: Alcohol, physical activity and other risk factors for colorectal cancer: a prospective study. Br J Cancer 55: , 1987 (6/) Hirayama T: Association between alcohol consumption and cancer of the sigmoid colon: observations from a Japanese cohort study [see comment citation in Medline]. Lancet 2: , 1989 (62) Stemmermann GN, Nomura AM, Chyou PH, et al: Prospective study of alcohol intake and large bowel cancer. Dig Dis Sci 35: , 1990 (63) Carstensen JM, Bygren LO, Hatschek T: Cancer incidence among Swedish brewery workers. Int J Cancer 45: , 1990 (64) Bjelke E: Epidemiologic studies of cancer of the stomach, colon, and rectum; with special emphasis on the role of diet. Scand J Gastroenterol Suppl 31:1-235,1974 (65) Williams RR, Horm JW: Association of cancer sites with tobacco and alcohol consumption and socioeconomic status of patients: interview study from the Third National Cancer Survey. J Natl Cancer Inst 58: , 1977 (66) Dean G, MacLennan R, McLoughlin H, et al: Causes of death of blue-collar workers at a Dublin brewery, Br J Cancer 40: , 1979 (67) Freudenheim JL, Graham S, Marshall JR, et al: Lifetime alcohol intake and nskof rectal cancer in western New York. Nutr Cancer 13: , 1990 (68) Longnecker MP: A case-control study of alcoholic beverage consumption in relation to risk of cancer of the right colon and rectum in men. Cancer Causes Control 1:5-14, 1990 (69) Kune S, Kune GA, Watson LF: Case-control study of alcoholic beverages as etiological factors: the Melbourne Colorectal Cancer Study. Nutr Cancer 9:43-56, 1987 (70) Tuyns AJ, Pequignot G, Gignoux M, et al: Cancers of the digestive tract, alcohol and tobacco. Int J Cancer 30:9-11, 1982 (7/) Ward K, Moriarty KJ, O'Neill S, et al: Alcohol and colorectal cancer. Gut 24:A981,1983 (72) Kabat GC, Howson CP, Wynder EL: Beer consumption and rectal cancer. Int J Epidemiol 15: , 1986 (73) Newcomb PA, Storer BE, Marcus PM: Cancer of the large bowel in women in relation to alcohol consumption: a case-control study in Wisconsin (United States). Cancer Causes Control 4: , 1993 (74) Stemmermann GN, Heilbrun LK, Nomura AM: Association of diet and other factors with adenomatous polyps of the large bowel: a prospective autopsy study. Am J Clin Nutr 47: , 1988 (75) Kono S, Dceda N, Yanai F, et al: Alcoholic beverages and adenomatous polyps of the sigmoid colon: a study of male self-defence officials in Japan. Int J Epidemiol 19: , 1990 (76) Cope GF, Wyatt JI, Pinder IF, et al: Alcohol consumption in patients with colorectal adenomatous polyps. Gut 32:70-72, 1991 (77) Kikendall JW, Bowen PE, Burgess MB, et al: Cigarettes and alcohol as independent risk factors for colonic adenomas. Gastroenterology 97: , 1989 (78) Kune GA, Kune S, Read A, et al: Colorectal polyps, diet, alcohol, and family history of colorectal cancer: a case-control study Nutr Cancer 16:25-30, 1991 (79) Honjo S, Kono S, Shinchi K, et al: Cigarette smoking, alcohol use and adenomatous polyps of the sigmoid colon. Jpn J Cancer Res 83: , 1992 (SO) Sandier RS, Lyles CM, McAuliffe C, et al: Cigarette smoking, alcohol, and the nsk of colorectal adenomas. Gastroenterology 104: , 1993 (81) Holden C: Alcoholism and the medical cost crunch. Science 235: ,1987 (82) Maree KA, van der Westhuzen J, Metz J: Interrelationship between serum concentrations of methionine, vitamin B12 and folate. Int J Vit Nutr Res 59: , 1989 (83) Selhub J, Jacques PF, Wilson PW, et al: Vitamin status and intake as primary determinants of homocysteinemia in an elderly population [see comment citation in Medline]. JAMA 270: , 1993 (84) Talley NJ, Chute CG, Larson DE, et al: Risk for colorectal adenocarcinoma in pernicious anemia. A population-based cohort study [see comment citation in Medline]. Ann Intern Med 111: , 1989 (85) Zeisel SH: "Vitamin-like" molecules. (A) Choline, chap 25. In Modern Nutrition in Health and Disease (Shils ME, Young VR, eds), 7th ed. Philadelphia: Lea & Febiger, 1988, pp (86) Mamett LJ: Aspirin and the potential role of prostaglandins in colon cancer Cancer Res 52: , 1992 (87) National Research Council: Diet and Health: Implications for Reducing Chronic Disease Risk. Washington, DC: Natl Acad Press, 1989 (88) Stampfer MJ, Malinow MR, Willett WC, et al: A prospective study of plasma homocyst(e)ine and risk of myocardial infarction in US physicians. JAMA 268: , 1992 Notes Supported by Public Health Service (PHS) grant CA55075 from the National Cancer Institute, National Institutes of Health (NIH), Department of Health and Human Services (DHHS); by PHS grant HL35464 from the National Heart, Lung, and Blood Institute, NIH, DHHS; and by Special Institution Grant No. 18 from the American Cancer Society. G. A. Colditz is supported by a Faculty Research Award (FRA-398) from the American Cancer Society. We are indebted to Mira Koyfman, Mildred Wolff, Elizabeth Frost-Hawes, Kerry Pillsworth, and Jill Arnold for expert help. Manuscript received June 30, 1994; revised October 24, 1994; accepted November 15, Journal of the National Cancer Institute, Vol. 87, No. 4, February 15, 1995 ARTICLES 273